Integrations

Connect an AI agent (MCP)

Manage ECARD from Claude, ChatGPT or another AI agent, with the same rights as you.

An AI agent (Claude, ChatGPT, Copilot…) can read and manage your ECARD account on your behalf: "add Marie Dupont to the Paris group", "which signatures does the sales team get?", "create a QR code to our new offer", "how many card scans this month?". It goes through ECARD's MCP server (Model Context Protocol, the open standard for connecting AI agents to applications).

The agent never has more rights than the person who authorized it: a manager limited to two spaces stays limited to those two spaces.

Before you start

  • The feature is in beta: ECARD opens it organization by organization. If you do not see the AI agents section in your settings, ask support to open it.
  • You need an agent that accepts custom MCP connectors (remote MCP server with OAuth sign-in).
  • You must be an Owner, Admin or Manager of the organization.

Enable AI agents

  1. The Owner or an Admin opens Organization settings → Security → AI agents (MCP).
  2. They switch on Allow AI agents. It is off by default; switching it off later cuts every connected agent immediately.
  3. Copy the MCP server address shown: https://mcp.ecardnfc.fr/mcp.

Connect your agent

  1. In your agent, add a custom connector (or "MCP server") and paste the address https://mcp.ecardnfc.fr/mcp.
  2. The agent opens an ECARD page: sign in with your usual account.
  3. On the authorization screen, tick the organizations concerned, then pick for each domain (users, signatures, QR codes, cards, campaigns, statistics…) None, Read or Write.
  4. Click Allow. The agent is connected: ask it for instance "What can you do on my ECARD account?".

If your agent does not offer new tools after an ECARD update, start a new conversation or reconnect the connector.

What the agent can do

DomainReadWrite
Userslist, full record, custom attributescreate, edit, contact details, photo, archive, restore
Groups and tagslistcreate, rename, assign users
Email signaturestemplates, a user's signatures, previewcreate (from blocks), duplicate, replace the design, targeting, switch off, switch back on
QR codeslist and scanscreate, change destination, switch off
NFC cardscards, orders and trackingmode, linked user, report lost
Profile pages, contact files (VCF), campaigns, walletstemplates, targeting, resultscreate or duplicate a template, targeting, visibility, visuals, publishing, archiving
Contactslist, recordcreate, edit, archive
Statisticsviews, scans, clicks, collected contacts—
Settingsdefaultspage visibility, default modes
Activity logwho did what and when—

The agent cannot: order or pay, invite admins, send bulk emails, delete permanently, export contacts, edit the design of an existing page, contact file or wallet pass (it can create a new one; for a signature it can replace its design).

A template the agent creates or duplicates applies to nobody until it is targeted: nothing changes for your users by surprise. It stays editable in the dashboard's editors.

Example requests

  • "List the users of the Marketing group who have no photo."
  • "Put the Lyon office address on Paul's and Léa's records."
  • "Start a campaign with this visual from December 1st to 15th for the Sales group."
  • "Show me a preview of Marie's signature."
  • "Compare card scans between September and October."
  • "Add a link to our website to the signature, shown without the https://" (the agent uses modifiers, here {{user.website|noprotocol}}).

Security

  • Same rights as you, reduced to the domains and levels ticked at authorization, and checked again on every action: if your role changes or you leave the organization, the agent loses access at once.
  • Ceiling per role: in Security → AI agents → What agents may do, an owner or admin sets, for admins and managers, the highest level per domain (None, Read, Write) and can forbid them to switch off or archive (signatures, QR codes, pages, campaigns, wallet passes, users, contacts, cards reported lost). It applies at once to agents already connected; the owner is never limited.
  • Confirmation: before archiving, unpublishing or reporting a card lost, the agent must get your agreement. When your application supports it, ECARD asks you directly.
  • Access log: every agent action, reads included, is listed in Security → AI agents, and every change appears in the activity log with the agent mentioned.
  • Revocation: remove an agent at any time from the list of connected agents.
  • The data the agent reads is sent to the AI provider you use: check that it complies with your internal policy.

Troubleshooting

  • The AI agents section does not show. The feature is not opened for your organization yet: contact support.
  • "The request expired". Start the connection again from your agent: the authorization link is only valid for a few minutes.
  • No organization to tick. Your role does not allow management, or the owner or an admin has not enabled AI agents.
  • "This agent was not granted…". The requested domain was not ticked at authorization: reconnect the agent and grant that access.
  • The agent only sees part of the data. Expected for a manager: only their spaces are reachable.

See also Integrations and Users and roles.

On this page